Analysis of the nonsteroidal anti-inflammatory drug literature for potential developmental toxicity in rats and rabbits.

Abstract

Nonsteroidal anti-inflammatory drugs (NSAIDs) are among the most commonly prescribed to pregnant women. Some case-control studies have linked the NSAIDs aspirin and indomethacin with a risk of congenital abnormalities and low birthweight. High doses of aspirin produce developmental toxicity in rats (e.g., gastroschisis/umbilical hernia, diaphragmatic hernia [DH]) when administered during sensitive windows of development. Unlike other NSAIDs, aspirin irreversibly inhibits cyclooxygenases (COXs) 1 and 2. Hence, the developmental toxicity seen in rats after exposure to aspirin may be due to the irreversible inhibition of COX-1 and/or COX-2. If so, other NSAIDs, which act through a reversible inhibition of COX, may produce a weak developmental toxicity signal or no developmental toxicity signal when tested in preclinical models. To investigate this relationship, a comprehensive analysis of the NSAID developmental toxicity literature was undertaken to determine whether NSAIDs other than aspirin induce developmental anomalies similar to those elicited by aspirin. Developmental toxicity studies were identified through literature searches of PubMed and TOXNET, and pregnancy outcome data were extracted and tabulated. By using a set of defined criteria, each study was evaluated for quality and assigned to one of five tiers. The relation between certain malformations and NSAID treatment was analyzed for the best studies (tiers 1-4) by using concurrent control data (Mantel-Haenszel and permutation tests) and by combining the concurrent control data with historical control data (chi2 test and permutation tests). A qualitative analysis of these data led to a focus on three types of malformations: DH, ventricular septal defects (VSDs), and midline defects (MDs). In rats, the incidences of VSD and MD were increased among fetuses treated with NSAIDs when compared with the concurrent controls. The extent of the increase was attenuated when the data from the aspirin studies were excluded from the analysis. There were no qualifying (i.e., tiers 1-4) aspirin studies conducted in rabbits, but the incidences of the three defects were increased over control incidences among non-aspirin NSAID-treated animals. Statistical analysis of these data was subsequently conducted. When tiers 1-4 were combined and compared with concurrent controls plus the most appropriate historical control database, the strongest associations were between NSAID treatment and VSD in rats, VSD in rabbits, and MD in rabbits. There also was some suggestion of an association between NSAID treatment and DH in rabbits. This analysis of the non-clinical NSAID literature demonstrated a possible association between exposure to NSAIDs and developmental anomalies. The anomalies were similar for aspirin and for other NSAIDs, but effects occurred at a much lower incidence with non-aspirin NSAIDs than previously reported with aspirin. Such a finding is consistent with the concept that reversible inhibition of COX-1 and/or COX-2 by other NSAIDs would produce weaker developmental toxicity signals than aspirin. However, there were limitations of the evaluated studies: (1) there were very few robust International Conference on Harmonization-compliant studies conducted with NSAIDs in the published literature; (2) many of the studies were conducted at doses well below the maximum tolerated dose (MTD), where effects are rarely seen; and (3) numerous studies were conducted above the MTD, where reduced numbers of fetuses hampered detection of low-incidence findings. Although weak associations were observed, these limitations prevented us from definitively determining the presence or absence of a developmental toxicity signal from the existing body of NSAID data. Further exploration of this hypothesis will require assessing the potential association in animal models by using dose levels centered around the MTD.